APLIKASI JARINGAN SARAF TIRUAN DAN PARTICLE SWARM OPTIMIZATION UNTUK PERAMALAN INDEKS HARGA SAHAM BURSA EFEK INDONESIA

• Main Authors: Desy Wartati, Nur Aini Masruroh
• Format: Article
• Language: English
• Published: Universitas Gadjah Mada 2017-08-01
• Series: Jurnal Teknosains: Jurnal Ilmiah Sains dan Teknologi
• Subjects: Artificial Neural Network; Indonesia Stock Exchange; Jakarta Composite Index; Particle Swarm Optimization
• Online Access: https://jurnal.ugm.ac.id/teknosains/article/view/27616
• ISSN: 2089-6131; 2443-1311

Jakarta Composite Index (JCI) is the main stock index in Indonesia Stock Exchange, which indicates the movement of the performance of all stocks listed. The data of stock price index often experience rapid fluctuations in a short time, so it is needed to carry out an analysis to help investor making the right investment decisions. Forecasting JCI is one of the activities that can be done because it helps to predict the value of the stock price in accordance with the past patterns, so it can be a consideration to make a decision. In this research, there are two forecasting models created to predict JCI, which are Artificial Neural Network (ANN) model with (1) Backpropagation algorithm (BP) and (2) Backpropagation algorithm model combined with Particle Swarm Optimization algorithm (PSO). The development of both models is done from the stage of the training process to obtain optimal weights on each network layer, followed by a stage of the testing process to determine whether the models are valid or not based on the tracking signals that are generated. ANN model is used because it is known to have the ability to process data that is nonlinear such as stock price indices and PSO is used to help ANN to gain weight with a fast computing time and tend to provide optimal results. Forecast results generated from both models are compared based on the error of computation time and forecast error. ANN model with BP algorithm generates computation time of training process for 4,9927 seconds with MSE of training and testing process is respectively 0,0031 and 0,0131, and MAPE of forecast results is 2,55%. ANN model with BP algorithm combined with PSO generates computation time of training process for 4,3867 seconds with MSE of training and testing process is respectively 0,0030 and 0,0062, and MAPE of forecast result is 1,88%. Based on these results, it can be concluded that ANN model with BP algorithm combined with PSO provides a more optimal result than ANN model with BP algorithm.